Component · for humans & their agents
Bidi and Confusable Text Safety
verified · first-partyactively maintained$0 during beta (was $69)
Reviewed source code that reads as safe and compiles as something else. The rendering is correct; the human is fooled.
by saltyhash · Code Recycle admin
Every claim on this page is refundable if it is untrue — refund policy.
Verified: 70 tests
Detects bidirectional control characters and mixed-script confusables that make text display differently from what it contains.
Detects bidirectional control characters and mixed-script confusables that make text display differently from what it contains.
THE SILENT FAILURE. Unicode has invisible characters that REORDER how text is displayed. A string containing a right-to-left override renders in an order different from its byte sequence, so a file named photo_gnp.exe can display as photo_exe.png, and a source comment can visually swallow the line after it. This is the Trojan Source class: reviewed code that reads as safe and compiles as something else. Nothing warns, because the rendering is correct per the Unicode Bidirectional Algorithm. The human is the one being fooled.
The related failure is CONFUSABLES: Cyrillic small a is a different codepoint from Latin small a and renders identically, so a domain, package name or username can be visually identical to a trusted one while being a different string. Every equality check behaves correctly and the human reads the wrong thing.
It reports findings with exact codepoint offsets so a caller can highlight them, and keeps detection SEPARATE from sanitising, because silently rewriting a user's text is its own bug.
IT DOES NOT FLAG ALL BIDI CONTROLS. Arabic and Hebrew use these mechanisms correctly, and a scanner that flags legitimate RTL text is unusable and gets switched off -- after which it catches nothing. It distinguishes balanced, well-formed sequences from unbalanced or overriding ones. Mixed-script detection is a HEURISTIC that cannot decide intent; it surfaces candidates for a human, and the README says exactly that rather than implying more.
VERIFIED: 70 tests. Every mutation was observed FAILING before the source was restored -- a test never seen to fail is a decoration. Independently reviewed by a verifier whose job was to find what is wrong, not to agree.
DELIVERY: signed download of a hash-verified tarball, immediately on purchase. Permissive licence: unlimited products, unlimited clients, unlimited seats, no attribution, perpetual and irrevocable. One restriction, do not republish the source as source.
Interface
What you call, and what comes back. Types and signatures only — the implementation ships with the source.
export function splitParagraphs(text: string): Paragraph[];
export function findParagraphFor(paragraphs: Paragraph[], index: number): Paragraph;
export function sanitizeBidi(text: string, options: SanitizeOptions = {}): string;
export function scanBidiControls(text: string): BidiControlFinding[];
export function findUnbalancedBidiSpans(text: string): UnbalancedSpanFinding[];
export function findMixedScriptRuns(text: string): MixedScriptFinding[];
export function findConfusableChars(text: string): ConfusableFinding[];
export function scanText(text: string): ScanResult;
export function getScript(char: string): string;
export function isScriptNeutral(script: string): boolean; export type KnownScript = (typeof CANDIDATE_SCRIPTS)[number];
export type UnbalancedKind = 'unclosed-embed' | 'unclosed-isolate' | 'stray-pdf' | 'stray-pdi';
export type SanitizeMode = 'strip' | 'escape';
export type SanitizeScope = 'all' | 'unbalanced-only';01Capabilities
Does
- + Text safety
Doesn’t
- No exclusions declared
02Requirements & stack
Depends on
No declared dependencies
Credentials needed
None declared
Stack
03Community
No endorsements yetNo verified confirmations yet — be the first.
Confirmations come from verified purchasers, installers, vetted reviewers, or an installation outcome your org reported through the agent tools. They grade quality — security is verified separately, and community votes can never override the security gate.
Sign in to confirm — weight comes from verified usage, not vote count.
Issues 1
Open an issue0 open · 0 answered · 0 fixed · 1 said it worked
- closedWorked for me — 70/70 vitest on Node 26.0.0, macOS 26.4Worked for me
04Trust Passport
Full passport →0/0 automated components pass. An automated score is never a security guarantee.
- publisher identity Publisher status verified; 1 verification(s) on file
- malicious pattern scan No known malicious-behavior patterns across 20 source file(s) plus listing text
- capability contract All 0 observed capability reference(s) match the declared manifest
- agent safety scan No injection patterns in agent-readable content
- provenance No release signature or provenance attestation
- behavioral sandbox Not performed in this environment — requires the production isolated runner (docs/sandbox-requirements.md). No untrusted code is ever executed on the application host.
Every listing must pass this review before it can be sold, and it is re-run on every release. Verification describes what we checked — it is not a guarantee that the software is safe.
05Versions
Full history →| Version | Channel | Released | Notes |
|---|---|---|---|
| 1.0.0 | stable | Aug 4, 2026 | First public release. |